|
Activity Number:
|
224
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #302034 |
|
Title:
|
Approximate Score Confidence Intervals for a Beta-binomial Proportion
|
|
Author(s):
|
Michael E. Schuckers*+
|
|
Companies:
|
St. Lawrence University
|
|
Address:
|
211 Valentine Hall, Canton, NY, 13617,
|
|
Keywords:
|
rate estimation ; false accept ; biometric identification ; false reject
|
|
Abstract:
|
We consider an adjustment to the Beta-binomial proportion similar to that made by Agresti and Coull for binomial proportions. Traditional methods for making confidence intervals for a Beta-binomial proportion tend to perform poorly when the intra-individual correlation is large. In addition, the likelihood of the product Beta-binomial is not tractable for making most classic confidence intervals. To overcome this, we use an approximate score test where we substitute the estimated variance based on the likelihood function with the estimated variance of the estimator assuming the null hypothesis holds. A confidence interval is then derived by solving the resulting quadratic equation. Both the original and a simplified version are considered. We then compare these approaches to traditional interval for a Beta-binomial proportion via Monte Carlo simulation. The application that motivates this work is the estimation of false accept and false reject error rates for biometric identification devices.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2004 program |
